Persistent neural activity has been observed in a wide range of brain regions and has been implicated in functions ranging from information storage and processing to motor control. Deficits in persistent neural activity in working memory areas of the brain have been suggested as a core feature of schizophrenia. The proposed work seeks to reveal the neural mechanisms underlying persistent neural activity by computational modeling of a model system exhibiting persistent neural activity, the goldfish oculomotor neural integrator. The oculomotor neural integrator receives velocity-coded eye movement commands and converts these into signals that control the position of the eyes. In the absence of velocity commands, neurons in the integrator maintain a steady rate of firing for tens of seconds. Patients with impaired neural integrators are unable to maintain a steady gaze and have deficits in eye tracking behavior and ocular reflexes. Previous models of the oculomotor system have neglected important features that have made them unable to be tested explicitly by experiment. Using a novel framework that allows data to be directly incorporated, an experimentally constrained and verifiable model of the goldfish oculomotor neural integrator will be constructed. The model will be used to analyze network and cellular contributions to persistent neural activity. The contributions of synaptic excitation, synaptic inhibition, and intrinsic neuronal excitability will be assessed by modeling recent anatomical and pharmacological manipulations of persistent neural activity in the system. Preliminary modeling at the network level suggests that recurrent interactions between cells are mediated by a bistable dendritic process that is hypothesized to be a dendritic plateau potential. A network model with dendritic branching structures and voltage-sensitive synaptic and intrinsic conductances will be constructed to test the hypothesis that voltage-dependent dendritic properties increase the robustness of the network to perturbations. The model will be constrained by intracellular recordings in slice and in vivo and will be compared to dendritic imaging experiments currently being conducted in the consultants'laboratories. By producing an experimentally constrained and verifiable model in a well-characterized system, this work promises to reveal core mechanisms by which persistent neural activity is generated.